Models

Alibaba's flagship hosted Qwen model, a very large MoE serving the top of the Qwen3.5 line. Excellent multilingual quality, strong reasoning and coding, and a large context window. First-party hosted, so pricing is fixed rather than routed across providers. A capable frontier option, especially for Chinese-English workloads, when you…

OpenAI's flagship, a frontier reasoning model with excellent coding, math and tool use and a large context window. Configurable reasoning effort lets you trade latency for depth on hard problems. First-party hosted with high uptime, priced at the top of the market. The model to reach for on genuinely difficult work.

Google's flagship Gemini, with frontier reasoning, native multimodality (text, image, audio, video) and a million-token context window. Fast for its size and backed by Google's infrastructure, it's a top pick for complex agents, long-document analysis and multimodal pipelines. First-party hosted at a premium price point.

Mistral's flagship, a strong general-purpose model with excellent multilingual coverage and reliable function calling. A good European-hosted option for production assistants and agents that need solid reasoning without going all the way to the most expensive frontier tiers. Available from a few providers with steady uptime; priced for…

Cohere's enterprise flagship, tuned for retrieval-augmented generation, grounded answers with citations, and multilingual business workloads. Strong tool use and a long context window, with the deployment and compliance options enterprises tend to need. Priced for production RAG rather than cheap bulk jobs.

Anthropic's balanced mid-flagship, widely used for coding and agentic work. It combines strong reasoning and excellent instruction-following with very reliable tool use, and it's a default backbone for many coding agents. First-party hosted with top-tier uptime, priced for production rather than high-volume bulk jobs.

xAI's frontier model, strong on reasoning, coding and real-time knowledge, with a large context window and fast tool use. First-party hosted with high uptime. A capable general assistant and agent backbone that trades a premium price for top-tier reasoning.

Anthropic's most capable model, built for the hardest reasoning, long-horizon agentic tasks and large-codebase work. Best-in-class instruction-following and tool use, first-party hosted with top-tier uptime. It sits at the very top of the price range, reserved for work where quality matters more than cost.